Assessing the link between changes in landscape and desertification in the chambal river basin using machine learning and remote sensing

This study aims to analyze the linkage between landscape change and dessertification over the last three decades in the Chambal River Basin (CRB), India, from 1990 to 2020, Machine Learning (ML) based Random Forest (RF) algorithm was implemented using remote sensing data to generate Land Use and Land Cover (LULC) maps, enabling landscape based change detection analysis to understand the impact of anthropogenic and natural drivers on desertification. The Google Earth Engine (GEE) platform was utilised for data processing and analysis, leveraging the Landsat series of satellite data. Overall accuracy for the years 1990, 1999, 2011, and 2020 was reported to be 84%, 84%, 82%, and 86% respectively. The intensity of drought was analysed using the Standardised Precipitation-Evapotranspiration Index (SPEI) across different accumulation scales (3, 6, 9, and 12 months) from 1990 to 2020. Correlation analysis between LULC dynamics and multi-temporal SPEI values revealed moderate negative associations in several periods, indicating that areas experiencing intense land transformation, particularly conversion of vegetation and agricultural lands, corresponded with lower moisture availability and higher drought stress. This suggests that anthropogenic land alterations have amplified the region’s vulnerability to climatic variability. The findings from the CRB indicate that 21% of the land cover area changed between 1990 and 2020. Furthermore, this study highlights that ML algorithms implemented through GEE offer an effective and efficient platform for analysing large datasets and conducting classification studies. The strong relationship observed between SPEI and LULC change underscores their combined usefulness in providing evidence for desertification in large areas, supporting future planning efforts.